The back propagation neural network, or "BPN", is an extremely useful neural network algorithm. The BPN "learns" a very general class of mappings which are usually represented as functions from R.sup.n to R.sup.m. Theoretically a 3-layer BPN can learn almost any map, but in practice the application of the BPN has been limited due to the enormous amount of computer time required for the training process.